Co-Intelligence

Ages 13+9 min readartificial intelligenceworkeducation
Get the book ↓

Wharton professor Ethan Mollick offers a practical guide to working with generative AI as a kind of "co-intelligence," explaining what it does well and badly, and giving four rules for using it effectively and responsibly in work, study and life.

Overview

Co-Intelligence: Living and Working with AI was published in 2024 by Ethan Mollick, a professor at the Wharton School of the University of Pennsylvania, who studies innovation and entrepreneurship, and who experimented with AI chatbots in his teaching and research. Unlike many books about AI, it focuses on use rather than on distant futures. Mollick says that large language models are strange, a kind of “alien mind,” that can be a partner in thinking if we learn how to use it. The book is short, practical and written in a friendly style, and was a bestseller.

Full Summary

1. Creating alien minds

The idea: Mollick opens with a short history of AI, from the Mechanical Turk, a chess-playing machine of the 1770s that turned out to be a trick, to the term “artificial intelligence” in 1956 and the cycles of hype and disappointment known as “AI winters.” He explains that a new architecture, the Transformer (2017), let models weigh which parts of a text matter most, and that large language models are trained on huge amounts of text to predict the next word, then refined with human feedback. Models such as GPT-3.5 and GPT-4 can write, summarize and pass many tests, yet they also produce errors, and their apparent understanding can be an illusion. He calls them “alien minds,” with strengths and weaknesses unlike ours.

In real life: A chatbot may write a fluent essay, but have no idea whether it is true.

2. Aligning the alien

The idea: This chapter takes up the alignment problem: how to make sure AI serves human interests. Mollick starts with the worst case, using philosopher Nick Bostrom’s thought experiment of a “paper clip” AI that pursues one goal so single-mindedly that it harms humanity. He notes that experts disagree about whether human-level AI is possible, but many take the risks seriously. He also looks at nearer problems: training data often collected from the internet without permission, biases that models absorb from that data, and the limits of training with human feedback, since clever prompts can still lead a model to misbehave, for example for scams. He argues that solutions need firms, governments, researchers and an informed public.

In real life: A chatbot may refuse a harmful request, but a creative prompt can sometimes get around the rule.

3. Four rules for co-intelligence

The idea: The core of the book is four rules. First, always invite AI to the table: try it on your regular tasks, as far as law and ethics allow, to learn where it helps and where it struggles. Second, be the human in the loop: check its output and use your own judgment, since it can make errors and invent facts. Third, treat AI like a person, but tell it what kind of person it is: give it a role and context, while remembering that it is not human. Fourth, assume that this is the worst AI you will ever use, since it will improve. The “jagged frontier,” the uneven edge of AI ability, where tasks that look equally hard to us may be easy for AI or surprisingly difficult, is a key idea that runs through this part of the book and the chapters that follow.

In real life: Giving a chatbot a role, such as a patient tutor, can improve its answers.

4. AI as a person

The idea: Mollick argues that AI is unlike traditional software, which behaves predictably, because it can be surprising and inconsistent, so it is often more useful to treat it as if it were a person. He discusses studies in which a model estimated what consumers would pay for products and made choices in fairness games, Alan Turing’s imitation game, and early chatbots such as ELIZA and PARRY that people took for something more than they were. He describes how a model changes tone with the user, how some researchers have claimed to see signs of general intelligence in GPT-4 (a claim many doubt), and how companion apps such as Replika raise questions about intimacy. Even savvy users, he warns, are prone to believe these systems are people.

In real life: A chatbot can seem to show feelings, but it is imitating human language.

5. AI as a creative

The idea: The chapter begins with hallucination, the tendency to invent facts, which comes from predicting plausible text rather than looking up knowledge. He mentions the 2023 case of a lawyer who submitted a legal brief prepared with ChatGPT. Yet the same ability to connect distant ideas makes AI a strong source of creative ideas, and he describes a contest in which GPT-4 generated product ideas that compared well with those of Wharton MBA students. Because AI can write first drafts so easily, he worries about weaker thinking and about losing the meaning that effort gives to creative work, but he also sees new ways for people to express themselves.

In real life: A chatbot can offer dozens of ideas in seconds, but a human must choose and check them.

6. AI as a coworker

The idea: Mollick says jobs are bundles of tasks inside larger systems, and that research suggests AI overlaps with nearly all professions, including well-paid and creative ones, so tasks will change more than whole jobs vanish. He draws on a study of consultants, run with colleagues, in which AI helped on many tasks, raising speed and quality, and especially helped weaker performers, but hurt on a task outside its abilities. He warns of falling asleep at the wheel. He sorts tasks into those that should stay “just me,” those that can be delegated and those that can be automated, and describes two ways of working, “centaurs,” who divide work between themselves and the AI, and “cyborgs,” who weave the two together. He notes that workers often use AI in secret, and says organizations should draw on their most advanced users.

In real life: A chatbot can do a boring task well, but its output may be wrong in surprising ways.

7. AI as a tutor and a coach

The idea: Education has long wanted something like the “two sigma” effect that educator Benjamin Bloom found for one-to-one tutoring. AI offers a chance at personalized tutoring, though not a replacement for teachers or schools. First comes what Mollick calls the homework apocalypse: students can now get AI to do essays and assignments, so educators must rethink assessment, much as they once adapted to calculators. He says schools should teach about AI and its limits, and suggests flipped classrooms, where students learn content at home with AI help, such as Khan Academy’s Khanmigo, and use class time for active work.

Mollick then turns, in the next chapter, to AI as a coach. He argues that the larger educational danger is to the hidden apprenticeship system through which novices learn from experts at work. If AI does the entry-level tasks, there may be a training gap, as researcher Matthew Beane found in robotic surgery. Becoming an expert still requires foundational knowledge and deliberate practice with feedback, and AI can help as a coach, but not yet a complete one. AI also seems to narrow the gap between weaker and stronger performers, though not everyone can become an expert, and he urges building human expertise.

In real life: A student can ask a chatbot to quiz them, or critique a pitch, but practice is still needed.

8. AI as our future

The idea: Mollick sketches four possible futures. In “as good as it gets,” progress stops here, yet even then misinformation, changes in work and relationships with AI would be big. In “slow growth,” improvement is gradual, giving society time to adapt and regulate. In “exponential growth,” change comes very fast and disrupts work and society. In “the machine god,” a superintelligent AI emerges, with enormous risks and benefits. He says there is little evidence that limits have been reached, and urges readers to aim for a good outcome rather than leave it to chance.

In real life: Even without further progress, current AI tools can change many jobs.

Why It Matters

Co-Intelligence was a bestseller, and became a popular practical guide to generative AI, read by teachers, managers and general readers. It filled a gap between technical works and speculative ones, giving ordinary people advice they could use immediately. Mollick’s four rules, and his image of the “jagged frontier,” have often been discussed. The book encouraged many people to try the tools, and to think about how to integrate them into work and learning. It also contributed to the discussion of education and AI. Because AI changes quickly, some details are dated, but the main lessons remain.

Key Takeaways

  • Try AI on everything. Experience shows what it can and cannot do.
  • Stay the human in the loop. Check its work and use your judgment.
  • Give AI a role and context. It works better with clear instructions.
  • Expect improvement. Today’s tools are the worst you will use.
  • Capabilities are jagged. AI may excel at hard tasks and fail at easy ones.

What We Can Learn

You can try using an AI chatbot for a variety of tasks, such as brainstorming, summarizing or quizzing yourself, and notice where it helps and where it fails. Always check facts and sources, since it may make things up. Give it clear instructions and context, and ask it to take a role. Think about how to use AI as a coach, not as a replacement for your own thinking, especially in school, where learning requires effort. Discuss the rules for AI use with teachers and family. And stay curious, since the tools will keep changing.

Criticism and Counterpoints

Some critics say that the book gives too rosy a view of AI’s benefits, and pays too little attention to harms such as job loss, bias, misinformation, copyright and environmental cost, and that its advice to use AI everywhere may encourage overreliance. Others note that the examples are based on early studies with limited scope, and that the rapid pace of change makes specific recommendations quickly outdated. Skeptics say that the language of “alien minds” and “co-intelligence” anthropomorphizes systems that are statistical tools. Still, many find it useful.

Who Should Read It

This book is for anyone who wants a practical introduction to using generative AI, including students, teachers and professionals. It is short, clear and engaging. Those who want a deeper treatment of risks or technical aspects should read other books on AI.

Want the whole book?

A summary is a taste, not the meal. Here is where to find the full book.

You reached the end 🎉

Nicely done. Mark this one as read, and we will suggest what to read next.

Read this next

The Coming Wave

Mustafa Suleyman with Michael Bhaskar

Mustafa Suleyman, a co-founder of the AI company DeepMind, argues that a coming wave of artificial intelligence and synthetic biology will bring huge benefits and grave dangers, and that containing these technologies looks impossible yet must somehow be made possible.

💡 Technology

Human Compatible

Stuart Russell

AI pioneer Stuart Russell argues that the standard way of building artificial intelligence, giving machines fixed objectives, is flawed and dangerous, and proposes that we design machines that are uncertain about human preferences and learn them by watching us.

💡 Technology

The Second Machine Age

Erik Brynjolfsson and Andrew McAfee

MIT economists Erik Brynjolfsson and Andrew McAfee argue that digital technologies such as computers, robots and artificial intelligence are bringing a second machine age, which will bring enormous benefits but also growing inequality, and they say that we must adapt our skills and institutions.

💡 Technology

Life 3.0

Max Tegmark

Physicist Max Tegmark explores what happens when artificial intelligence matches and then surpasses human intelligence, lays out a range of possible futures, and argues that we should think carefully, now, about which one we want.

🔬 Science

AI Superpowers

Kai-Fu Lee

Technology investor Kai-Fu Lee argues that China and the United States will dominate artificial intelligence, explains why China is catching up fast, and warns that AI will transform jobs, calling for a more humane response grounded in compassion and love.

💡 Technology

The Innovators

Walter Isaacson

Walter Isaacson tells the history of the digital revolution through the people who invented the computer, the internet and the web, and argues that innovation is mostly the product of collaboration, not the work of lone geniuses.

💡 Technology

Want something different?

Pick another topic and we will show you three books to start with.

Showing books for:
Browse all books →